Alexander Kuzmenkov
8629b29083
complain about unstable perf test queries
...
We don't want to introduce more of them. The old ones should be all marked by now.
2021-05-12 12:48:13 +03:00
Alexander Kuzmenkov
0b990c9519
Merge pull request #23514 from ClickHouse/aku/perf-test-cpu-model
...
upload cpu model to perf test db
2021-04-23 09:25:51 +03:00
Alexander Kuzmenkov
3f855cb714
more
2021-04-22 20:18:21 +03:00
Alexander Kuzmenkov
6a1414663f
upload cpu model to perf test db
2021-04-22 19:48:32 +03:00
Azat Khuzhin
6fba6f7ddd
Disable min_bytes_to_use_mmap_io by default
...
Reading files using mmap() does not have any significant benefits over
plain read() [1].
[1]: https://gist.github.com/azat/3d6c8d82bdd91e7a38d997fd6bcfd574
And not only it does not have significant benefits, it also has some
issues, due to max_server_memory_usage (default to 90% of available
RAM), since when you read files with mmap() eventually process RSS may
exceed max_server_memory_usage, and in this case any allocation will
fail (with "Memory limit exceeded (total)") error (yes kernel will
unload pages, but likely it will happens after queries will starting to
fail), like in this test [2].
[2]: https://gist.github.com/azat/4813489828162e6c2ce131963c6a1acb
TL;DR;
Note that there was also an idea to take those mmap()'ed regions in
memory tracking (#23211 ), but there are some drawbacks (since accounting
mmap() is tricky, first of all you need to account only once per inode
for file and plus kernel can unload some pages and those memory will not
be used by the server anymore).
And as an adddition to #23211 there was #23212 , that adds
max_bytes_to_use_mmap_io, but since mmap is not a subject for memory
accounting there is no need in it.
2021-04-20 00:25:16 +03:00
Alexander Kuzmenkov
2a4bcb6e3f
Merge pull request #23159 from ClickHouse/aku/merge-fusecount
...
merging sumCount fusion PR #21337
2021-04-19 16:47:13 +03:00
Azat Khuzhin
7083d126f7
Disable postgresql_port in perf tests
...
To avoid port overlaps
2021-04-17 15:28:09 +03:00
Alexander Kuzmenkov
2489b6af96
cleanup
2021-04-15 19:40:49 +03:00
Alexander Tokmakov
9925110713
rename other occurrances
2021-03-29 23:04:50 +03:00
Alexander Kuzmenkov
b0284f20c3
Merge branch 'master' into aku/laglead
2021-03-22 21:16:27 +03:00
Alexander Kuzmenkov
5ae5134147
Merge pull request #21846 from ClickHouse/aku/perf-test-error
...
more robust error handling in perf test
2021-03-22 17:18:37 +03:00
Alexander Kuzmenkov
1beba597ca
Merge remote-tracking branch 'origin/master' into HEAD
2021-03-19 02:11:08 +03:00
Alexander Kuzmenkov
6aa9039f7d
float frames and lag/lead_in_frame
2021-03-19 02:05:43 +03:00
Alexander Kuzmenkov
8ab778e26a
Merge pull request #21754 from ClickHouse/aku/fuzzer-ci-sources
...
make the fuzzer use sources from the CI
2021-03-18 18:02:13 +03:00
Alexander Kuzmenkov
f2ac17cfba
Update compare.sh
2021-03-18 17:59:49 +03:00
Alexander Kuzmenkov
065faf1318
more robust error handling in perf test
2021-03-18 01:32:55 +03:00
Alexander Kuzmenkov
a78b234a6d
fixes
2021-03-15 21:45:57 +03:00
Denis Glazachev
6f4e883d8d
Factor out settings
...
Enable parallel insert select
Reduce the amount of data to process
2021-03-06 15:09:45 +04:00
Nikolai Kochetov
a195e783d2
Try fix perftests.
2021-03-04 13:25:49 +03:00
Alexander Kuzmenkov
f169be740a
Merge pull request #21381 from ClickHouse/aku/perf-negative
...
fix a rare false negative in perf tests
2021-03-03 15:55:14 +03:00
Nikolai Kochetov
019513d869
Try fix perftests.
2021-03-03 13:42:44 +03:00
Alexander Kuzmenkov
08148e062f
Fix a rare false negative in perf tests
2021-03-02 19:21:30 +03:00
Azat Khuzhin
9a26409147
Fix performance tests (by avoid sharding status file for right and left server)
...
Since cp -al (hard links):
2021.02.21 01:09:09.991771 [ 243 ] {} <Information> StatusFile: Status file right/db/status already exists - unclean restart. Contents:
PID: 241
Started at: 2021-02-21 01:09:09
Revision: 54448
2021.02.21 01:09:09.992007 [ 243 ] {} <Error> Application: DB::Exception: Cannot lock file right/db/status. Another server instance in same directory is already running.
2021-02-21 10:56:48 +03:00
Alexander Kuzmenkov
c18749a704
fix a bug and add some tests
2021-02-09 17:44:04 +03:00
Amos Bird
a2ceca3fa1
Fix recursive propagation
2021-01-26 10:29:12 +08:00
Azat Khuzhin
de7909211d
Try using top_level_domains from the patched performance archive
...
If there is top_level_domains list in the upstream/master, use from the
patched version (this is required to run all performance tests for
upstream/master in the PR).
2021-01-02 17:27:03 +03:00
Azat Khuzhin
469b631ca6
Fix performance comparison
...
In #18113 the top_level_domains list copying was moved into
docker/packager/binary/build.sh, this was done to avoid symlinks (since
Dockerfile cannot dereference them).
But the patch was wrong, since it copied into the root (/), which is not
included into performance.tgz and also compare.sh was not modified.
This wasn't showed up with CI checks since the docker image wasn't
updated and it still included that top_level_domains, once it was
modified the image was updated and it became broken.
Cc: @akuzm
2021-01-02 17:09:20 +03:00
Azat Khuzhin
95c35b318e
Avoid using symlinks for top_level_domains
...
Move the main copy into the tests/config and ship it to the perf tests
via the perf package.
2020-12-15 21:58:46 +03:00
Alexander Kuzmenkov
5f424f7c5c
Update zzz-perf-comparison-tweaks-config.xml
2020-12-14 20:08:32 +03:00
Alexander Kuzmenkov
e7f93b5563
Merge pull request #18058 from azat/perf-test-zk-port-fix
...
Remove test_keeper_server in perf tests
2020-12-14 17:54:47 +03:00
Alexander Kuzmenkov
2edabd5519
Update zzz-perf-comparison-tweaks-config.xml
2020-12-14 17:45:58 +03:00
Azat Khuzhin
653ddab63e
Remove test_keeper_server in perf tests
...
To avoid port overlaps for right and left server (anyway it is not
used).
2020-12-14 10:59:01 +03:00
Azat Khuzhin
6f2c559398
Convert top_level_domains_lists.xml into regular file
...
docker/test/performance-comparison/config/config.d is not allowed to has
symlinks, since it is copied with the COPY docker command.
2020-12-10 21:01:17 +03:00
Azat Khuzhin
0f1d9f3a6f
Put top_level_domains into the docker/test/performance-comparison/config
...
Since in docker/test/performance-comparison there is Dockerfile that is
runned from that directory, so it cannot COPY anything outside of it.
2020-12-09 21:08:31 +03:00
Azat Khuzhin
8a4b29495b
perf: override top_level_domains_path via command line arguments
2020-12-09 21:08:31 +03:00
Azat Khuzhin
7b0c720922
Add top_level_domains into perf tests
...
And change Dockerfile to trigger image rebuild
2020-12-09 21:08:31 +03:00
Azat Khuzhin
79e6348282
perf: load overrides last of all
2020-12-09 07:31:54 +03:00
Azat Khuzhin
d33b0ad33b
Use interserver_http_port remove=remove over auto-port
...
Revert "perf: fix port conflict for tcp_with_proxy_port"
This reverts commit 89b3566824e5c0a87b6310fc4b6f9d300d2e17db.
2020-12-09 00:11:12 +03:00
Azat Khuzhin
75f9640fc2
perf: fix port conflict for tcp_with_proxy_port
...
Simply use zero port to do auto allocation, anyway it is not used in
perf tests.
2020-12-08 23:29:38 +03:00
Azat Khuzhin
f1ac8775db
perf: override config directives via command line over sed by config.xml
2020-12-08 21:25:03 +03:00
Azat Khuzhin
3696916f44
perf: add setuptools into docker image (to built clikchouse-driver)
2020-11-27 07:20:41 +03:00
Azat Khuzhin
04a93f18e8
perf: fail on unknown settings
2020-11-27 00:16:41 +03:00
Azat Khuzhin
a591886740
perf: remove outdated comments from perf.py
...
Refs: https://github.com/ClickHouse/ClickHouse/pull/17255#discussion_r528822709
2020-11-27 00:16:41 +03:00
Azat Khuzhin
a38628d967
perf: use the latest available clickhouse-driver (from git)
...
Since later setting_is_important will be used, but there is no release
with this feature included.
2020-11-27 00:16:41 +03:00
myrrc
fbb0e6e6aa
Merge remote-tracking branch 'upstream/master' into improvement/diff-types-in-avg-weighted
2020-11-24 16:04:17 +03:00
Alexander Kuzmenkov
a543c8e468
Add an option to use existing tables to perf.py
2020-11-23 21:05:54 +03:00
Azat Khuzhin
d676e2f5c5
Pass settings for perf tests via protocol over SET query
...
Thus they will be applied on reconnects, since clickhouse-driver may
implicitly do it if the connection has been failed.
2020-11-21 14:02:33 +03:00
Azat Khuzhin
3ddb0ac53e
Configure logging for perf test runner
2020-11-21 14:02:33 +03:00
Azat Khuzhin
f6d16ee43e
Force clickhouse-driver >= 1.1.5 for settings-as-strings support for perf-tests image
2020-11-21 14:02:33 +03:00
Alexander Kuzmenkov
b0e660651a
cleanup
2020-11-16 19:09:58 +03:00